When companies produce items, personnel typically examine them for flaws. Consider a scratch on a smartphone’s screen or a structural flaw in raw steel that could cause issues later in the production process. Landing AI, founded by leading AI expert Andrew Ng, previously of Google and Baidu, aims to utilize artificial intelligence to detect these imperfections, and the company has now introduced a new visual inspection system known as LandingLens.

“We are announcing LandingLens, a complete visual inspection platform designed to assist manufacturers in developing and implementing AI-powered visual inspection systems,” Ng explained to TechCrunch.

He stated that the company’s objective is to integrate AI into manufacturing operations, but simply adapting his previous work from Google and Baidu wasn’t feasible. This was due to differing consumer applications and, importantly, the limited amount of data typically available in a manufacturing environment.

The challenge is further compounded by the fact that each manufacturing environment is distinct, and a universal approach cannot be reliably applied across all industries. Consequently, Landing AI developed a versatile toolkit that each company can tailor to the specific demands of its production procedures.

Ng explained that to make this advanced technology accessible to clients and apply AI to visual inspection, his company has created a user-friendly interface that guides companies through a structured process for training models to meet their individual inspection requirements.

The process involves providing images of both acceptable and defective products. However, this is more complex than it appears, as even experienced human inspectors may not always agree on what constitutes a defect.

Manufacturers create what is termed a “defect book,” where inspection specialists collaborate to define the appearance of defects through images and resolve any disagreements that arise. This entire process is managed through the LandingLens interface.

After inspectors establish a consistent set of criteria, they can utilize the Model Iteration Module to train and refine models until the AI consistently and accurately identifies defects. The software generates reports on model performance, allowing customers to make necessary adjustments based on the provided data.

Ng emphasized that his company is providing advanced software to address a significant challenge for manufacturers. “The primary obstacle [for them] is developing the deep learning algorithm, essentially the machine learning software. They can assess a part and determine if it meets standards or is defective, and our platform facilitates this process,” he stated.

He believes this technology has the potential to fundamentally change manufacturing processes in the future. “I believe deep learning is on the verge of revolutionizing inspection procedures, which is a critical step. Inspection serves as the final safeguard against quality defects in manufacturing, and I am pleased to release this platform to help manufacturers improve inspection accuracy,” he said.